There was a study reported somewhere in the modeling media within the last few years doing strength tests on wood joints glued up with all the usual types of modeling products. The results confirmed my general suspicions, that the wood will fail before the glue joint fails, no matter what the wood or glue. Even crappy ARF hot-melt joints are hard to pull apart if they're joined well to begin with.

That said, I've had CA joints fail because of brittleness of CA relative to epoxy. If a joint is likely to be exposed to shock stress, use epoxy. Same thing with vibration stress, as in the firewall. Plus epoxy can be filleted with a popsicle stick or latex-gloved fingertip if you thicken it with colloidal silica, which is a marvelous additive (keeps the epoxy from running or sagging while curing).

Structural joints that will end up being covered over with film or glass are less critical in terms of joint strength than those that depend on glue strength alone. The monokote/ultracote skin is a wonderful stress distributor.

Hardworking joints like firewall, landing gear mounting plate, joining wing halves, wing-mount blocks--these get epoxied and cured overnight. Gluing in tailpieces, I've used both CA and epoxy, and yet to have a tail-joint of mine fail--BUT I have seen them come loose on other models, so you do need to be sure you get complete adhesion, fill all cracks and crannies. If a tailpiece needs bracing into alignment, obviously epoxy is the way to go, at least 12-minute cure to give you time to line everything up.